RICC, June 2005

Controlling the Evolution of a Simulated Hurricane through

Optimal Perturbations: Initial Experiments Using a 4-D

Variational Analysis SystemR. N. Hoffman, C. Grassotti, J. M.

Henderson, S. M. Leidner, G. Modica, and T. Nehrkorn

Atmospheric and Environmental Research

Lexington, MA

RICC, June 2005

Thanks• Supported by

NIAC

– NASA Institute for Advanced Concepts

• Tools & data

– MM5/4d-VAR

– NCAR/NCEP gridded data

H. Iniki 1992 (NWS image)

RICC, June 2005

Today’s talk

• Experiments to control hurricanes

• A different approach to weather control–not just hurricanes

• Based on the sensitivity of the atmosphere

• The same reason why it is so difficult to predict the weather!

RICC, June 2005

Theoretical basis

• The earth’s atmosphere is chaotic

• Chaos implies a finite predictability time limit no matter how well the atmosphere is observed and modeled

• Chaos also implies sensitivity to small perturbations

• A series of small but precise perturbations might control the evolution of the atmosphere

RICC, June 2005

Objectives of our project• Develop a method to calculate

the atmospheric perturbations needed to control a hurricane

• Quantify the size of the perturbations needed to do this

• Estimate the requirements of a weather control system

RICC, June 2005

Current NWP operational practice

• NWP centers have developed forecast techniques that capitalize on the sensitivity of the atmosphere

1. 4D variational data assimilation

2. Generation of ensembles

3. Adaptive observations

RICC, June 2005

Current Practice 1: 4D variational data assimilation• 4D-Var fits all available observations

during a time window (6 or 12 hours) with a model forecast

• The fit to the observations is balanced against the fit to the a priori or first guess from a previous forecast

• We use a variant of 4D-Var in our experiments

RICC, June 2005

Why hurricanes?

• Public interest: Threat to life and property

• History: Project Stormfury (1963)

• Sensitive to initial conditions

• MM5/4d-VAR: Available tools

RICC, June 2005

Our Case Study Hurricane Iniki (1992)• Landfall at Kauai at 0130 UTC

12 September

• Hurricane Iniki from 0600 UTC 11 September to 1200 UTC 12 September 1992 is shown in the following movie

RICC, June 2005

Iniki Simulation

750-hPa Relative Humidity

RICC, June 2005

Determination of perturbations

• Optimal control theory

• 4d-Var methodology baseline

• Modified control vector: temperature only

• Refined cost function: property damage

RICC, June 2005

Mesoscale model• The MM5 computation grid is 200 by

200, with a 20 km grid spacing, and ten layers in the vertical

• Physics are either– Simplified parameterizations of the

boundary layer, cumulus convection, stratiform cloud, and radiative transfer; or

– Enhanced parameterizations of these physical processes and a multi-layer soil model

RICC, June 2005

4D variational data assimilation• 4D-Var adjusts initial conditions

to fit all available observations during a 6 or 12 hour time window

• The fit to the observations is balanced against the fit to the a priori or first guess from a previous forecast

• We use a variant of 4D-Var in our experiments

RICC, June 2005

Standard 4D-Var cost function

• J is the cost function

• P is the perturbed forecast

• G is the goal

– G is the target at t=T and the initial unperturbed state at t=0

• S is a set of scales

– S depends only on variable and level

• x is temperature or a wind component

• i, j, and k range over all the grid points

J = ∑xijkt [(Pxijk(t)–Gxijk(t))/Sxk]2

RICC, June 2005

• Optimal is defined as simultaneously minimizing both the goal mismatch and the size of the initial perturbation as measured by the sum of squared differences

‘Optimal’ Defined

RICC, June 2005

Modified control vector

• Control vector can be restricted by variable and by geographic region

– Temperature only

– Locations far from the eye wall

RICC, June 2005

Refined cost function

• JD = ∑ijt Dij(t) Cij

• C is the replacement cost

• D is the fractional wind damage– D = 0.5 [1 + cos(π(V1-V)/(V1-V0))]

•D=0 for V<V0 = 25 m/s

•D=1 for V>V1 = 90 m/s

– Evaluated every 15 min. for hours 4–6

RICC, June 2005

Minimization

RICC, June 2005

Experiments• Hurricane Andrew MM5 simulations

starting at 00 UTC 24 Aug 1992

• Initial conditions from an earlier 6 h forecast; NCEP reanalysis; bogus vortex

• 4d-Var over 6 h (ending 06 UTC 24 Aug); 20 km grid; temperature increments only; simple physics

• Simulations for unperturbed vs. controlled; 20 km simple physics vs. 7 km enhanced physics

RICC, June 2005

Initial conditionsProperty values

RICC, June 2005

Temperature perturbations

RICC, June 2005

Surface wind field evolution

RICC, June 2005

Control vector sensitivity

RICC, June 2005

Temperature increments

RICC, June 2005

Temperature perturbations(controlled minus unperturbed)

RICC, June 2005

Time evolutionof perturbations

RICC, June 2005

Surface wind field evolution

Unperturbed

Controlled

RICC, June 2005

Time evolution 00 UTC

RICC, June 2005

Time evolution 06 UTC

RICC, June 2005

Time evolution 12 UTC

RICC, June 2005

Time evolution 18 UTC

RICC, June 2005

High resolution

RICC, June 2005

Summary

• Perturbations calculated by 4d-Var

• Control path, intensity of simulated hurricane

• Power requirements are huge– Higher resolution, longer lead times

may help

– Very large scale SSP could meet the requirements

RICC, June 2005

Use in Forecasting

• 4dVAR can assess the likelihood of a specific event that requires immediate action, such as damaging winds along the Hudson Valley

• Exigent, adj:

– 1 : requiring immediate aid or action

– 2 : requiring or calling for much

RICC, June 2005

Background cost function• Exigent forecasting: Normal

NWP Jb should be used. Related to the probability of the initial conditions.

• Weather control: Jb should be replaced with the cost in terms of available resources of generating the perturbations.

RICC, June 2005

Hurricane WxMod

• Energetics

– Biodegradable oil

– Pump cold water up to the surface

• Dynamic perturbations

– Stormfury: cloud seeding

– Space based heating

RICC, June 2005

Space based heating

• Solar reflectors: bright spots on the night side and shadows on the day side

• Space solar power (SSP): microwave downlink could provide a tunable atmospheric heat source

RICC, June 2005 Space solar power

NASA artwork by Pat Rawlings/SAIC

RICC, June 2005

Microwave spectrum

• Water and oxygen are the main gaseous absorbers– H2O lines at 22, 183 GHz

– H2O continuum

– O2 lines at 60, 118 GHz

• Frequency and bandwidth control the heating profile

RICC, June 2005

Microwave heating rates

RICC, June 2005

Power requirements

• Heating rates calculated for 1500 W/m2

• Equal to 6 GW/(2 km)2

• Current experiments require similar heating rates over an area 100s times larger

• Longer lead times, higher resolution will reduce these requirements significantly.

• By changing the storm’s environment at longer lead times can we prevent its forming, track, or intensity.

RICC, June 2005

The future• More realistic experiments:

resolution, physics, perturbations

• Future advances in several disciplines will lead to weather control capabilities– The first experiments will not be space

based control of landfalling hurricanes!

• Can legal and ethical questions be answered

RICC, June 2005

More complicating factors• The control must be effected at

significant time lags

• The difficulty of effecting control

• The problem of defining “optimal”

– For inhabitants of New Orleans, eliminating a hurricane threat to that city may take precedence over all else

RICC, June 2005

Future WxMod

• Improved models, observations, and assimilation systems will advance to the point where forecasts are:– much improved, and

– include an estimate of uncertainty

• Thus allowing advance knowledge that a change should be detectable in particular cases

RICC, June 2005

end

• Contact: – rhoffman at aer dot com

– www.niac.usra.edu

• Background: – R. N. Hoffman. Controlling the global

weather. Bull. Am. Meteorol. Soc., 83(2):241--248, Feb. 2002.